Revenue-Side Vs Spending-Side Fiscal Adjustments Evidence from Italian Municipalities

Transcription

1 Revenue-Side Vs Spending-Side Fiscal Adjustments Evidence from Italian Municipalities Luigi Marattin - University of Bologna Tommaso Nannicini Bocconi University Francesco Porcelli SOSE SpA University of Rome- La Sapienza 9 May 2017

2 SUMMARY 1. MOTIVATION AND OVERVIEW 2. LITERATURE REVIEW 3. SOME THEORETICAL INSIGHTS 4. EMPIRICAL ANALYSIS 4.1. Preliminary analysis 4.2. The data 4.3. The model and main results 5. CONCLUSIONS AND NEXT STEPS

3 OUR QUESTION We are concerned with the composition of the fiscal adjustment at the local level. When receving a trasfer cut from the central government, does a municipality meet its budget constraint by acting on the revenue side or the expenditure side?

4 1. Motivation and overview The reaction of local authorities to central governments spending cuts has been only marginally explored so far both by the theoretical and empirical literature This paper aims at closing some of the literature gaps providing a theoretical framework and an empirical analysis showing how local authorities react to a contraction of intergovernmental grants.

5 We sketch a theoretical framework whose aim is simply to gain some insights on the conditions under which local authorities are more prone to increase local taxes rather than reducing local expenditure to cope with the grants contraction No connection with the empirical model. The empirical analysis relies on new and unique dataset of Italian municipalities over a period of three years from 2010 to 2012.

6 Italy provides a good case-study for the empirical analysis due to the intense program of intergovernmental grants reduction implemented since 2010 (8,43 billions from 2010 to 2015) Moreover, Italian data are particularly useful because of the presence of a discontinuity in the allocation of the grants reduction: municipalities below 5000 inhabitants have been permanently exempted from the spending cut of 2.5 bln implemented by the central government in 2010 (DL 78/2010).

7 Our identification strategy allows us to obtain standard difference in discontinuities estimates of the Italian municipalities fiscal adjustment behaviour. By doing so, we provide a good measure of the causal relationship between central and local fiscal decisions. Final results show that Italian municipalities reacted to the contraction of inter-governmental grants mainly by increasing local taxes rather than by reducing local expenditure

8 Macro Composition (of fiscal adjustment) matters. Alesina et al (2015, JIE): on average, the output costs of expenditurebased fiscal adjustments are close to zero. Inter-governmental Skidmore (1999, Public Choice) on US: imposing tax limitations tends to increase unconstrained revenue sources rather than reduce spending Gong and Zue (2002, JEDC): theoretical framework On Italy: Gennari and Messina (2014, Int.Tax.Pub.Fin): flypaper effect Political economy 2. Related literature Political cycle: Brender (2003, J.PubEco), Brender and Drazen (2005, JME), Drazen and Eslava (2010, J.Dev.Eco).

9 Y Federal Government (FG) moves G t and τ t+1 Its budget constraint is: G t + Γ t = τ Y t Y t K Local Government (LG) moves g t and τ t+1 Its budget constraint is: g t = Γ t + τ k t K t Households (HS) chooses C t to maximize: U = (C t, G t, g t ) Its budget constraint is: K t+1 = (1 τ t Y )Y t K t C t + 1 τ t K K t Timing First move FG Second move LG Third move HS 3. Theoretical framework

10 Theoretical framework Optimization process Households choose consumption C t and capital K t+1 (under their resource constraint) in order to maximize the discounted sum of instantaneous utility function U C t = βu Y C t+1 1 τ t+1 Y t+1 + (1 τ K K t+1 ) t+1 the optimum the opportunity cost of accumulating one unit of new capital (LHS) must be equal to the discounted net marginal benefit of future consumption (RHS). In turn, this latter is the sum of the net-of-taxation unit of capital and the net-of-taxation output produced by that unit. K The local government chooses local public spending g t and local tax instrument τ t+1 (under its budget constraint and FOCs) in order to maximize welfare function (=utility function) U C t+1 λ 2,t = U g t+1 K t+1 the trade-off facing the LG: if it increases local tax instrument τ K, the benefit is future marginal utility of local public spending (RHS), but the cost is the foregone future marginal utility of consumption. Y The federal government chooses federal public spending G t and federal tax instrument τ t+1 (under its budget constraint, LG FOC and HS FOC) in order to maximize welfare function (=utility function) λ 5,t U C t+1 Y t+1 K t+1 = U G t+1 at the optimum the marginal cost of increasing federal tax (LHS) must be equal to the marginal benefit of the corresponding increase in public spending or transfer to the local government.

11 Theoretical framework - Equilibrium The equilibrium is pictured by a vector of endogenous variables: [ K t+1, C t, G t, g t, τ Y t+1, τ K t+1, Γ t ] At the equilibrium: g Γ = τk K > 0 τy when the central government reduces transfers to local government (Γ), the latter reduces local public spending g. Such a reduction is greater the higher the steady-state level of local tax rate (τ K ) and the stock of capital (K). Re-arranging we ge: τ K = gτ Y KΓ τ Y KY τ K Γ = gτy K (KΓ τ Y KY) 2 when the central government reduces transfers to local government, the latter increases the local tax rate (τ K ). Such an increase is greater the higher the steady-state level of local public spending and the steady-state tax rate on income.

12 4. EMPIRICAL ANALYSIS 4.1. Preliminary Analysis 4.2. The data 4.3. The model and main results

13 art. 61 c. 11 DL 112/2008 art. 47 c. 8 DL 66/2014 e art. 9 DL 16/2014 art. 1 c. 435 L 190/2014 art. 28 c. 7 DL 201/2011 art 14 c. 2 DL 78/2010 art. 16 c. 6 DL 95/2012 Billion euros I TAGLI DEI COMUNI 4.1. Preliminary analysis Since 2010 the Italian government has been conducting an intense program of spending cuts for municipalities (over 8000 elected local authories) TOTAL 8,6 billions (18% of current expenditure and 33% of 2007 capital expenditure) Source Ministry of Internal Affairs 2009 => art. 61 c. 11 DL 112/2008 (0,2 billions) 2011 => art 14 c. 2 DL 78/2010 (2,5 billions exemption of municipalities < 5000 inhabitants) 2012 => art. 28 c. 7 DL 201/2011 (1,4 billions) 2013 => art. 16 c. 6 DL 95/2012 (2,6 billions exemption municipalities affected by 2009 and 2012 the earthquake) 2014 => art. 47 c. 8 DL 66/2014 e art. 9 DL 16/2014 (0,7 billions) 2015 => art. 1 c. 435 DL 190/2014 (1,2 billions exemption municipalities affected by 2009 and 2012 the earthquake)

14 I TAGLI DEI COMUNI Allocation rules of spending cuts S i = β X i + ε i where: i = municipal index; S = per capita spending cuts; X = matrix of financial and environmental variables, β coefficients estimated with OLS Total spending cuts art. 61 c. 11 DL 112/2008 art 14 c. 2 DL 78/2010 art. 28 c. 7 DL 201/2011 art. 16 c. 6 DL 95/2012 art. 47 c. 8 DL 66/2014 e art. 9 DL 16/2014 art. 1 c. 435 L 190/2014 tax revenues current expenditure capital expenditure Income real estate value altitude zone (1=low - 5=high) degree of urbanization(1 = low - 3 = high) population n. of municipalites R-squared linear allotment criteria on spending or resources strong impact of population levels different criteria depending on the type of reduction Impact of population

15 milion euros I TAGLI DEI COMUNI Current expenditure composition 40,000 35,000 30,000 25,000 20,000 15,000 10,000 5, personnel expenditure Consumption Grants to other local gov. Interests other current expenditure Expnediture net of public transport and waste management. Only municipalities of normal regions 15 Source Ministry of Internal Affairs

16 no. of municipalities Municipalities by population above >500 >1000 >2000 >3000 >5000 >10000 >20000 >60000 > > > Special regions Normal regions , , Normal regions Special regions Source Ministry 16 of Internal Affairs, year 2014

17 Italian municipalities institutional framework 6700 municipalities (comuni ) [ around 600 > 15,000 inhabitants], ruled by a local government (giunta), headed by an elected mayor (sindaco ). Mayors: elected directly by residents (from1993); Plurality system for municipalities below inhab. Runoff system for municipalities above inhab. in power for five years and are subject to a two-term limit; power to nominate/dismiss members of the local government; competencies on environment, local transport, local police, culture and education; Some discretionary powers on how to raise fiscal revenue; supported by a coalition of parties There are two main party coalitions in Italy: centre-left and centre-right.

18 euro per capita Expenditure composition 3,000 2,500 2,000 1,500 1, population brackets Current expenditures Source Ministry of Internal Affairs, year 2014 Capital expenditures

19 Municipalities provides 10 main public services (billion euros): waste management (7.61) general admin. services (6.39) social services (4.67) complementary education services (3.57) public roads and transport (3.21) land management and planning (2.69) local police (2.64) nursery services (1.44) tax office (0.50) civil registry (0.55)

20 euros per capita Local Governments' Financing 2,500 2,000 1,500 1, population brackets taxes grants current grants capital debt Source Ministry 20 of Internal Affairs, year 2014

21 Local Governments' Financing Municipalities' revenues come from two main sources: transfers from upper levels of government (mainly from regional governments), from 2010 to abolition of central transfers => horizontal equalization system Taxes: Property Tax (ICI/IMU);Tax on Waste Disposal (TARSU/TARI); Other Taxes (Income tax and electricity surcharge). Fees for: general services (road and traffic, libraries, theatres and culture, burial services, etc.); other services (occupation of public spaces, public billboards, issues of certificates).

22 4.2. The data The (administrative) dataset contains yearly information at the municipality level about budget items (only normal regions => 6700 local authorities) Source: Italian Ministry of Internal Affairs, ISTAT (National Institute of Statistics); Italian Ministry of Economy and Finance Budget variables: Spending cuts current expenses, capital expenses (investments) (local) tax revenues, fees & tariffs, transfers Domestic stability pact target Merged with dataset containing political information: term limit, electoral outcomes, electoral cycle, turn out Merged with dataset containing a set of control variables: population by age, income, real estate market values Panel structure: from 2010 to 2012

23 Data snapshot Variable All sample Treatment = 1 (pop > 5000) Treatment = 0 (pop < 5000) Obs Mean Std. Dev. Obs Mean Std. Dev. Obs Mean Std. Dev. After treatment dummy Treatment dummy Spending cuts (dl 78/2010) ( per capita) Spending cuts (total) ( per capita) Current expenditure ( per capita) Capital expenditure ( per capita) Property tax ( per capita) Local income tax ( per capita) Fees (net of waste manag.) ( per capita) Total tax revenues ( per capita) Domestic stability pact (target) ( per capita) Population (0-2) % total pop Population (3-14) % total pop Population (15-65) % total pop Population (over 65) % total pop Real estate declared income ( per capita) Declared employment income ( per capita) Residential estate market values ( per sq meter) Electoral cycle discrete (0-4) Term limit dummy Electoral year dummy Turnout % margin of victory %

24 Euros per capita I TAGLI DEI COMUNI Analysis of the raw data 1000 Source Ministry of Internal Affairs Current expenditure Capital expenditure Property taxes Local income tax Fees Spending cuts 24 Expenditure net of public transport and waste management. Only municipalities of normal regions

25 4.3. The model and main results Y P = budget outcome = population size, P c = 5000 Y 1 = potential outcome with D = 1 (spending cuts) Y 0 = potential outcome with D = 0 (no spending cuts) After T 0 D = 1 if P P c D = 0 if P < P c Before T 0 D = 0 for all Main estimate of interest: after treatment effect = E[Y 1 Y 0 P = P c ]

26 Potetial outcome: Y(1) = spending cuts, Y(0) = no spending cuts The empirical strategy Post-treatment period E ( Y(1) P, t >= T0] γ 1 E ( Y(0) P, t >= T0] π γ 0 γ 0 = Conf. Effects post 2010 (no cuts) γ 1 = Conf. Effectcs post 2010 (cuts) π = Effect of spending cuts Pc = 5000 Population size observed not observed

27 Potetial outcome: Y(1) = spending cuts, Y(0) = no spending cuts The empirical strategy Pre-treatment period E ( Y(1) P, t < T0] γ 1 E ( Y(0) P, t < T0] π γ 0 γ 0 = Conf. Effects pre 2010 (no cuts) γ 1 = Conf. Effects pre 2010 (cuts) π = Effect of spending cuts Pc = 5000 Population size observed not observed

28 The empirical strategy Estimation methods Assumptions No time variation in the confounding discontinuity: γ 1 = γ 1 No interaction effect between treatment and confounding effects: γ 0 = γ 1 Diff-in-disc estimator identifies: E[Y1 - Y0 P = Pc] π = lim P Pc E Y P, t T o lim P Pc E Y P, t < T o lim P Pc E Y P, t T o lim P Pc E Y P, t < T o = E Y 1 Y 0 P = P c Estimation of the dierence of two discontinuities at Pc with: Local Linear Regression (LLR) Spline polynomial approximation

29 The empirical strategy LLR For given bandwidth, restrict the sample to cities in the interval P i [P c, P c + ] Y i = δ o + δ 1 P i + D i γ 0 + γ 1 P i + T i α o + α 1 P i + D i β 0 + β 1 P i + ε i Where D i is the treatment, T i = 1 after 2010, P i is the normalized population and standard errors are clustered at the city level β 0 identifies the after treatment effect Robusteness checks => multiple bandwithds

30 The empirical strategy spline polynomial approximation Flexible functional form to fit the relationship between Y and P on either side of P c in the entire sample p p k k Y i = (δ k P i ) + D i (γ k P i ) k=0 k=0 p p k k + T i (α k P i ) + D i (β k P i ) k=0 k=0 + ε i Where D i is the treatment, T i = 1 after 2010, P i is the normalized population and standard errors are clustered at the city level β 0 identifies the after treatment effect Robusteness checks => multiple p-order polynomials

31 Diff-in-disc jumps in spending cuts (T 1 T 0 ) Euros per capita (T 1 T 0 ) Regression function fit Sample average within bin 2th order global polynomial

32 Diff-in-disc jumps in current expenditure (T 1 T 0 ) Euros per capita (T 1 T 0 ) Regression function fit Sample average within bin 2th order global polynomial

33 Diff-in-disc jumps in total revenues (T 1 T 0 ) Euros per capita (T 1 T 0 ) Regression function fit Sample average within bin 2th order global polynomial

34 Total tax revenues (cut at 1000 inhab.) (1) (2) (3) (5) VARIABLES did llr llr lpr_2 After treatment effect 29.76*** 48.26*** 38.06*** 53.07*** Treatment effect *** *** *** *** T = *** 18.31** 44.66*** 27.09** Domestic Stability Pact (target) 0.116*** 0.116*** % pop *** 36.85*** % pop ** 17.10*** % pop *** 21.92*** % pop over *** 24.30*** Real estate declared income 0.210*** 0.209*** Declared employment income ** *** Residential estate market values *** *** Electoral cycle 26.54*** 26.89*** Electoral cycle sq *** *** Term limit *** *** Election year 17.47** 17.67** % turnout 0.368** 0.332* % margin of victory Constant 255.7*** 268.2*** -2,139*** -2,194*** Polynomia no first first second Observations 5,067 5,067 4,752 4,752 R-squared standard errors are clustered at the city level, *** p<0.01, ** p<0.05, * p<0.1

35 Total current expenditure (cut at 1000 inhab.) (1) (2) (3) (5) VARIABLES did llr llr lpr_2 After treatment effect * Treatment effect *** 34.34*** 18.77** T = *** ** Domestic Stability Pact (target) % pop % pop % pop * 19.29* % pop over * 21.79** Real estate declared income 0.137*** 0.139*** Declared employment income *** *** Residential estate market values 0.119*** 0.119*** Electoral cycle Electoral cycle sq Term limit Election year % turnout ** *** % margin of victory 0.466*** 0.469*** Constant 581.4*** 544.3*** -1,264-1,263 Polynomia no first first second Observations 5,168 5,168 4,843 4,843 R-squared standard errors are clustered at the city level, *** p<0.01, ** p<0.05, * p<0.1

36 Robustness checks - Total tax revenues (1) (2) (3) (4) VARIABLES lpr_2 cut 1000 lpr_3 cut 1000 lpr_3 cut 1500 lpr_3 cut 500 After treatment effect 53.07*** 50.11* 49.94** Treatment effect *** ** T = ** 49.05*** 30.59** Domestic Stability Pact (target) 0.116*** 0.117*** *** % pop *** 37.43*** 45.52*** 31.75*** % pop *** 17.66*** 21.39*** % pop *** 22.29*** 26.75*** % pop over *** 24.71*** 29.76*** Real estate declared income 0.209*** 0.207*** 0.203*** 0.196*** Declared employment income *** *** *** Residential estate market values *** *** *** *** Electoral cycle 26.89*** 26.52*** 28.46*** 22.64** Electoral cycle sq *** *** *** *** Term limit *** *** * ** Election year 17.67** 17.07** 19.65*** % turnout 0.332* 0.353** 0.353** % margin of victory ** Constant -2,194*** -2,257*** -2,707*** -1,294 Polynomia second third third third Observations 4,752 4,752 7,240 2,306 R-squared standard errors are clustered at the city level, *** p<0.01, ** p<0.05, * p<0.1

37 Robustness checks - Total current expenditure (5) (7) (8) (9) VARIABLES lpr_2 lpr_3 lpr_3 cut 1500 lpr_3 cut 500 After treatment effect Treatment effect *** T = ** Domestic Stability Pact (target) % pop *** % pop ** % pop * 19.72* 34.73*** % pop over ** 22.21** 38.22*** Real estate declared income 0.139*** 0.139*** 0.167*** *** Declared employment income *** *** *** *** Residential estate market values 0.119*** 0.119*** 0.105*** 0.120*** Electoral cycle Electoral cycle sq Term limit Election year % turnout *** *** *** ** % margin of victory 0.469*** 0.474*** 0.333*** 0.622*** Constant -1,263-1,320-2,861*** Polynomia second third third third Observations 4,843 4,843 7,375 2,350 R-squared standard errors are clustered at the city level, *** p<0.01, ** p<0.05, * p<0.1

38 Robustness checks - Total spending cuts (1) (2) (3) (5) (7) (8) (9) VARIABLES did cut 1000 llr cut 1000 llr cut 1000 lpr_2 cut 1000 lpr_3 cut 1000 lpr_3 cut 1500 lpr_3 cut 500 After treatment effect 32.19*** 29.16*** 29.27*** 21.87*** 16.94** 20.68*** Treatment effect 3.469*** 3.791*** 3.639*** 4.278*** 3.357*** 3.909*** 2.502* T = *** 46.87*** 46.01*** 52.03*** 54.30*** 53.19*** 58.15*** Domestic Stability Pact (target) *** *** *** *** *** % pop *** *** *** *** *** % pop *** *** *** *** *** % pop *** *** *** *** *** % pop over *** *** *** *** *** Real estate declared income *** *** *** *** *** Declared employment income * Residential estate market values *** *** *** *** Electoral cycle ** ** ** *** Electoral cycle sq ** 0.419** 0.410** 0.504*** Term limit Election year * % turnout *** *** *** *** *** % margin of victory Constant 1.334*** 1.408*** 489.8*** 496.4*** 500.4*** 469.0*** 634.7*** Polynomia no first first second third third third Observations 5,158 5,158 4,836 4,836 4,836 7,371 2,356 R-squared standard errors are clustered at the city level, *** p<0.01, ** p<0.05, * p<0.1

39 Diff-in-disc jumps - Revenues compostion (T 1 T 0 ) Regression function fit Regression function fit Sample average within bin 2th order global polynomial 0 Euros per capita (T 1 T 0 ) Euros per capita (T 1 T 0 ) Euros per capita (T 1 T 0 ) Euros per capita (T 1 T 0 ) Total tax revenues Property tax Sample average within bin 2th order global polynomial Local Income tax Regression function fit Fees (no waste manag.) Regression function fit Sample average within bin 2th order global polynomial Sample average within bin 2th order global polynomial

40 Diff-in-disc jumps - Revenues compostion Property tax Local income tax Fees VARIABLES did cut 1000 lpr_2 cut 1000 did cut 1000 lpr_2 cut 1000 Did cut 1000 lpr_2 cut 1000 After treatment effect 16.39*** 33.97** 4.152** *** 22.11** Treatment effect * *** 1.252* 2.806* *** *** T = *** 21.89** 14.22*** 16.60*** *** *** Domestic Stability Pact (target) *** % pop *** *** 8.798** % pop *** *** % pop *** *** % pop over *** *** Real estate declared income 0.173*** ** *** Declared employment income *** *** Residential estate market values *** *** *** Electoral cycle 30.13*** Electoral cycle sq *** Term limit *** Election year 19.59*** % turnout 0.791*** ** *** % margin of victory *** 0.157*** Constant 151.1*** -3,569*** 40.53*** 1,369*** 65.42*** Polynomia no second no second no second Observations 5,130 4,812 5,121 4,799 5,151 4,826 R-squared standard errors are clustered at the city level, *** p<0.01, ** p<0.05, * p<0.1

41 5. CONCLUSIONS AND NEXT STEPS Final results show that Italian municipalities reacted to the contraction of intergovernmental grants mainly by increasing local taxes rather than by reducing local expenditure Main policy implication: Transmission mechanism from central spending cuts to local spending in the presence of local fiscal autonomy Next steps: Validity tests Analysis of heterogeneous effects Discontinuity in 2012 and 2015 spending cuts (earthquakes exemptions)